Go God: The AI Agents Have Arrived

In the past 30 days, more than 3,000 AI tools have been launched. You may remember the iPhone advertisement “There’s an app for that”? Well now, there’s an AI for that, too.

And, unlike almost every human and corporation scrambling to get a brochure-ware website in 1999, most of them have incredible utility.

These past few weeks have been a really important time in AI, because it is starting to show how fast the recursion is. In other technologies and in other industrial breakthroughs, the recursive iterations took years. Think how long how long we had to wait to get from cell phone to smart phone, from the graphical web to democratised social media. Everything took years. But now, these incredibly innovative AI breakthroughs are being measured in weeks and days.

This has significant implications for investors and capital allocation because we’ve always had time to digest what is happening. Even though markets have always moved rapidly, the players and tools inside them didn’t. In addition AI is no longer the tool that big tech, builds and deploys upon us. It has been democratised, like media was, but it’s different because it is general in nature and omni-directional. We are in the middle of an intelligence explosion.

Generative AI is changing what’s possible so fast it is mind-boggling.

We’re All Coders Now

It used to be debatable on which coding language was most valuable - Python, Ruby, Java. Now…. it's English, or whichever language you prefer. Large Language Models (LLM's) haven't just given us an ability to get a zillion things done by asking ChatGPT.

Now, everyone can code. All we need to know is what to ask the machine to make for us.

I know what you're thinking: “That's fine Steve, I can ask ChatGPT to write me some computer code for a new app or site. But I wouldn't even know what to do with it, where to put it, or any of the implementation steps whatsoever.”

Refer above.

We ask it these questions too and it will give us a step-by-step process of exactly what to do and how to implement each step.

Well, that was what you’d do a few weeks ago. Now it’s even easier than that.

Now we can “go God Mode”. We can get Agent AIs to do all of this implementing autonomously.

AI Agents Have Arrived

Agent AIs take a user input, break it down into many small individual tasks with help of LLMs and work through them step-by-step, like any project manager would, only better. The Agent AIs use OpenAI to create, prioritize, and execute tasks. The main idea behind this system is that it actually creates its own tasks. Each new task it gives itself will be based on the result of previous tasks via a predefined objective which you have provided the Agent with.

The Agent then uses natural language processing (NLP) capabilities to iterate through the project based on the objective which has been set. This is seriously a revolution. It’s like having a personal employee or even a large team, with a PhD on every subject imaginable, at your disposal for free.

We Are The New Gods

Currently the user interfaces (there are already quite a few of these) are a bit clunky because they are mostly browser-based extensions. Of course, we can expect that to be fixed in a week or two as well. Here are a couple of Agent AIs you could experiment with:

• Baby AGI
• AgentGPT
• AutoGPT
• Auto GPT GUI
• God-Mode

Once you have a play with these tools, you’re bound to have an OMG moment. If the task has digital components it can do them automatically. But even when some real-world physical interaction is required, the Agents will go onto a digital forum and hire the people needed (say a freelancer or a delivery person) to act as their proxy to execute against this physical, real world requirement. All we need to do is provide them with the log ins, or payment facilities.

With generative AI coding, anyone will be able to simply express what they desire in natural language (in my case, English), transforming our ideas into reality. In this fashion, the tech will allow a doctor, lawyer, or small business owner to describe their desires, with the AI then automating code generation and task optimization. Or to manage any project imaginable to be done by an Agent instead of an employee.

This is vital because most software to date has been developed by coders, not end users, and it is why much of it rarely hits the mark. Starting now, our AI Agents will develop exactly the software we need for us and undertake to do complex tasks and projects on our behalf.  Now, one size fits one.

While Agent AIs have only emerged in the last few weeks, they’re developing at a breakneck pace. And now, Microsoft has gotten in on the act with Jarvis / HuggingGPT. As investors we now need to pay attention to what things can scale without much human or financial capital -- and which firms are about to vanish because of this new reality.